Adaptive training of neural networks for control of autonomous mobile robots

Research output: Chapter in Book/Report/Conference proceedingChapterAcademicpeer-review

Abstract

We present an adaptive training procedure for a spiking neural network, which is used for control of a mobile robot. Because of manufacturing tolerances, any hardware implementation of a spiking neural network has non-identical nodes, which limit the performance of the controller. The adaptive training procedure renders the input-output maps of these non-identical nodes practically identical, therewith recovering the controller performance. The key idea is to replace the nodes of the spiking neural network by small networks of synchronizing neurons that we call clusters. The networks (and interaction weights) are generated adaptively by minimizing the errors in the collective input-output behavior of the cluster relative to that of a known reference. By means of numerical simulations we show that our adaptive training procedure yields the desired results and, moreover, the generated networks are consistent over trials. Thus, our adaptive training procedure generates optimal network structures with desired collective input-output behavior.

LanguageEnglish
Title of host publicationSensing and Control for Autonomous Vehicles - Applications to Land, Water and Air Vehicles
EditorsT.I. Fossen, H. Nijmeijer, K.Y. Pettersen
PublisherSpringer
Pages387-405
Number of pages19
Volume474
ISBN (Print)9783319553719
DOIs
StatePublished - 2017
EventWorkshop on Sensing and Control for Autonomous Vehicles: Applications to Land, Water and Air Vehicles, 20-22 June 2017, Alesund, Norway - Alesund, Norway
Duration: 20 Jun 201722 Jun 2017

Publication series

NameLecture Notes in Control and Information Sciences
Volume474
ISSN (Print)0170-8643

Conference

ConferenceWorkshop on Sensing and Control for Autonomous Vehicles: Applications to Land, Water and Air Vehicles, 20-22 June 2017, Alesund, Norway
CountryNorway
CityAlesund
Period20/06/1722/06/17

Fingerprint

robot
neural network
hardware
performance
tolerance
manufacturing
simulation
interaction

Cite this

Steur, E., Vromen, T., & Nijmeijer, H. (2017). Adaptive training of neural networks for control of autonomous mobile robots. In T. I. Fossen, H. Nijmeijer, & K. Y. Pettersen (Eds.), Sensing and Control for Autonomous Vehicles - Applications to Land, Water and Air Vehicles (Vol. 474, pp. 387-405). (Lecture Notes in Control and Information Sciences; Vol. 474). Springer. DOI: 10.1007/978-3-319-55372-6_18
Steur, E. ; Vromen, T. ; Nijmeijer, H./ Adaptive training of neural networks for control of autonomous mobile robots. Sensing and Control for Autonomous Vehicles - Applications to Land, Water and Air Vehicles. editor / T.I. Fossen ; H. Nijmeijer ; K.Y. Pettersen. Vol. 474 Springer, 2017. pp. 387-405 (Lecture Notes in Control and Information Sciences).
@inbook{8c0ca8f0585c458cb4b3927c5605e6b8,
title = "Adaptive training of neural networks for control of autonomous mobile robots",
abstract = "We present an adaptive training procedure for a spiking neural network, which is used for control of a mobile robot. Because of manufacturing tolerances, any hardware implementation of a spiking neural network has non-identical nodes, which limit the performance of the controller. The adaptive training procedure renders the input-output maps of these non-identical nodes practically identical, therewith recovering the controller performance. The key idea is to replace the nodes of the spiking neural network by small networks of synchronizing neurons that we call clusters. The networks (and interaction weights) are generated adaptively by minimizing the errors in the collective input-output behavior of the cluster relative to that of a known reference. By means of numerical simulations we show that our adaptive training procedure yields the desired results and, moreover, the generated networks are consistent over trials. Thus, our adaptive training procedure generates optimal network structures with desired collective input-output behavior.",
author = "E. Steur and T. Vromen and H. Nijmeijer",
year = "2017",
doi = "10.1007/978-3-319-55372-6_18",
language = "English",
isbn = "9783319553719",
volume = "474",
series = "Lecture Notes in Control and Information Sciences",
publisher = "Springer",
pages = "387--405",
editor = "T.I. Fossen and H. Nijmeijer and K.Y. Pettersen",
booktitle = "Sensing and Control for Autonomous Vehicles - Applications to Land, Water and Air Vehicles",
address = "Germany",

}

Steur, E, Vromen, T & Nijmeijer, H 2017, Adaptive training of neural networks for control of autonomous mobile robots. in TI Fossen, H Nijmeijer & KY Pettersen (eds), Sensing and Control for Autonomous Vehicles - Applications to Land, Water and Air Vehicles. vol. 474, Lecture Notes in Control and Information Sciences, vol. 474, Springer, pp. 387-405, Workshop on Sensing and Control for Autonomous Vehicles: Applications to Land, Water and Air Vehicles, 20-22 June 2017, Alesund, Norway, Alesund, Norway, 20/06/17. DOI: 10.1007/978-3-319-55372-6_18

Adaptive training of neural networks for control of autonomous mobile robots. / Steur, E.; Vromen, T.; Nijmeijer, H.

Sensing and Control for Autonomous Vehicles - Applications to Land, Water and Air Vehicles. ed. / T.I. Fossen; H. Nijmeijer; K.Y. Pettersen. Vol. 474 Springer, 2017. p. 387-405 (Lecture Notes in Control and Information Sciences; Vol. 474).

Research output: Chapter in Book/Report/Conference proceedingChapterAcademicpeer-review

TY - CHAP

T1 - Adaptive training of neural networks for control of autonomous mobile robots

AU - Steur,E.

AU - Vromen,T.

AU - Nijmeijer,H.

PY - 2017

Y1 - 2017

N2 - We present an adaptive training procedure for a spiking neural network, which is used for control of a mobile robot. Because of manufacturing tolerances, any hardware implementation of a spiking neural network has non-identical nodes, which limit the performance of the controller. The adaptive training procedure renders the input-output maps of these non-identical nodes practically identical, therewith recovering the controller performance. The key idea is to replace the nodes of the spiking neural network by small networks of synchronizing neurons that we call clusters. The networks (and interaction weights) are generated adaptively by minimizing the errors in the collective input-output behavior of the cluster relative to that of a known reference. By means of numerical simulations we show that our adaptive training procedure yields the desired results and, moreover, the generated networks are consistent over trials. Thus, our adaptive training procedure generates optimal network structures with desired collective input-output behavior.

AB - We present an adaptive training procedure for a spiking neural network, which is used for control of a mobile robot. Because of manufacturing tolerances, any hardware implementation of a spiking neural network has non-identical nodes, which limit the performance of the controller. The adaptive training procedure renders the input-output maps of these non-identical nodes practically identical, therewith recovering the controller performance. The key idea is to replace the nodes of the spiking neural network by small networks of synchronizing neurons that we call clusters. The networks (and interaction weights) are generated adaptively by minimizing the errors in the collective input-output behavior of the cluster relative to that of a known reference. By means of numerical simulations we show that our adaptive training procedure yields the desired results and, moreover, the generated networks are consistent over trials. Thus, our adaptive training procedure generates optimal network structures with desired collective input-output behavior.

UR - http://www.scopus.com/inward/record.url?scp=85020478515&partnerID=8YFLogxK

U2 - 10.1007/978-3-319-55372-6_18

DO - 10.1007/978-3-319-55372-6_18

M3 - Chapter

SN - 9783319553719

VL - 474

T3 - Lecture Notes in Control and Information Sciences

SP - 387

EP - 405

BT - Sensing and Control for Autonomous Vehicles - Applications to Land, Water and Air Vehicles

PB - Springer

ER -

Steur E, Vromen T, Nijmeijer H. Adaptive training of neural networks for control of autonomous mobile robots. In Fossen TI, Nijmeijer H, Pettersen KY, editors, Sensing and Control for Autonomous Vehicles - Applications to Land, Water and Air Vehicles. Vol. 474. Springer. 2017. p. 387-405. (Lecture Notes in Control and Information Sciences). Available from, DOI: 10.1007/978-3-319-55372-6_18